Computer-Vision API from Streamlit

General
1

Hi iNaturalist team,

I maintain a small personal Streamlit app called HikeJournal for my own hike/photo field journal. It posts my own observations to iNaturalist using my own account/token and uses iNaturalist’s computer-vision suggestion endpoint only to pre-fill IDs for photos I personally upload.

The app was working from Streamlit Community Cloud, but recently the computer-vision suggestion request began returning 403 HTML/Cloudflare-style blocks from the hosted Streamlit environment. Normal posting still works.

I am not scraping or bulk-downloading data. Typical use is small personal batches of field photos, and I can keep requests to about 1 request/second or slower. The app now sends a User-Agent like:

HikeJournal/1.0 (personal field journal; contact: MYEMAIL (obfuscated for the forum post))

Is there a supported way to request access or allowlisting for this use case, or should I avoid using the CV suggestion endpoint from hosted infrastructure?

Thanks!

Tony

2

Have you registered your application? https://www.inaturalist.org/oauth/applications

3

the computer vision endpoint has never been officially available for arbitrary public use. i would guess this is probably iNaturalist staff finally cutting off undesired use.

4

Turns out 10 improving identifications for somebody like me is quite challenging haha! Soon I can get those 10 and register as an app!

5

have a look through ‘unknown’ observations at https://www.inaturalist.org/observations?identified=false; if you’re able to add basic IDs like ‘plant’ or ‘insect’ to some of these, those count as improving identifications

6

Thanks for this, that let me hit 10 and submit an application request. Since posting, I am not getting the 403 every time now. I did get the error towards the end of a batch, but I was able to immediately submit the rest, so I’m going to just further stagger my calls a bit to slow down as well.

Thanks to everybody who has chimed in!